Entering edit mode
> Date: Sun, 03 Jul 2005 10:13:29 +0000
> From: Carolyn Fitzsimmons <carolyn.fitzsimmons at="" imbim.uu.se="">
> Subject: Re: [BioC] duplicateCorrelation and design matrix
> To: bioconductor at stat.math.ethz.ch
>
> Hi Gordon, thanks for your reply. I have a few more questions:
>
> Quoting Gordon K Smyth <smyth at="" wehi.edu.au="">:
>
>> > Date: Thu, 30 Jun 2005 11:44:02 +0000
>> > From: Carolyn Fitzsimmons <carolyn.fitzsimmons at="" imbim.uu.se="">
>> > Subject: [BioC] duplicateCorrelation and design matrix
>> > To: Bioconductor list <bioconductor at="" stat.math.ethz.ch="">
>> >
>> > Hello,
>> >
>> > I need an explanation of how the design matrix influences the
consensus
>> > correlation of the duplicateCorrelation function when accounting
for
>> technical
>> > replicates. Here is my specific example:
>> >
>> > Design matrix:
>> >> design
>> > RJf RJm WLf WLm
>> > 1 0 0 0 1
>> > 2 0 0 0 1
>> > 3 0 0 0 1
>> > 4 0 0 0 1
>> > 5 0 0 0 1
>> > 6 0 0 0 1
>> > 7 0 0 0 1
>> > 8 0 0 0 1
>> > 9 0 0 1 0
>> > 10 0 0 1 0
>> > 11 0 0 1 0
>> > 12 0 0 1 0
>> > 13 0 0 1 0
>> > 14 0 0 1 0
>> > 15 0 0 1 0
>> > 16 0 0 1 0
>> > 17 0 1 0 0
>> > 18 0 1 0 0
>> > 19 0 1 0 0
>> > 20 0 1 0 0
>> > 21 0 1 0 0
>> > 22 0 1 0 0
>> > 23 0 1 0 0
>> > 24 0 1 0 0
>> > 25 1 0 0 0
>> > 26 1 0 0 0
>> > 27 1 0 0 0
>> > 28 1 0 0 0
>> > 29 1 0 0 0
>> > 30 1 0 0 0
>> > 31 1 0 0 0
>> > 32 1 0 0 0
>> > #
>> > each second slide is a replicate of the first (eg. 1 and 2 are
replicates,
>> then
>> > 3 and 4,... etc.). There are also 4 groups that I want to
compare, with 4
>> > individuals in each group (each duplicated). So I continue with
the
>> > duplicateCorrelation:
>> > #
>> >> cor <- duplicateCorrelation(Mmatrix_ny, design=design,
>> > +
>> >
>>
> block=c(1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13,
14,14,15,15,16,16))
>> >> cor$cor
>> > [1] -0.03060575
>> > #
>> > which is a pretty bad correlation so I probably should just use
the
>> technical
>> > replicates as biological replicates (the limma user guide says).
But in
>> > another comparison I want to put all the arrays in 2 groups, see
design
>> > matrix:
>> >> designWLRJ
>> > RJ WL
>> > 1 0 1
>> > 2 0 1
>> > 3 0 1
>> > 4 0 1
>> > 5 0 1
>> > 6 0 1
>> > 7 0 1
>> > 8 0 1
>> > 9 0 1
>> > 10 0 1
>> > 11 0 1
>> > 12 0 1
>> > 13 0 1
>> > 14 0 1
>> > 15 0 1
>> > 16 0 1
>> > 17 1 0
>> > 18 1 0
>> > 19 1 0
>> > 20 1 0
>> > 21 1 0
>> > 22 1 0
>> > 23 1 0
>> > 24 1 0
>> > 25 1 0
>> > 26 1 0
>> > 27 1 0
>> > 28 1 0
>> > 29 1 0
>> > 30 1 0
>> > 31 1 0
>> > 32 1 0
>> > #
>> > and then do the duplicateCorrelation function and get a different
>> correlation.
>> > #
>> >> corWLRJ <- duplicateCorrelation (Mmatrix_ny, design=designWLRJ,
>> > +
>> >
>>
> block=c(1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13,
14,14,15,15,16,16))
>> >> corWLRJ$cor
>> > [1] 0.01745252
>> > #
>> > Moreover when I compute the consensus correlation without using a
design
>> matrix
>> > I get 0.1073055. I know from looking through previous posts and
a lot of
>> help
>> > from Johan L. that the way the blocking is set up and using the
design
>> matrix
>> > in these situations is correct.
>>
>> You've used three different non-equivalent design matrices. No
more than one
>> of these can be
>> correct.
>
> But if I need to group the individuals differently to test for
differential
> expression between different groupings of individuals (i.e. between
> WLm/WLf/RJm/RJf and WL/RJ), the use of 2 different design matrixies
in the
> dupCorrelation function is warrented, yes?
No. Unless you have a good reason to do otherwise, set the full
design matrix and use
contrasts.fit() to group the individuals for differential expression
tests.
Gordon
>
>>
>> > So how is the consensus correlation actually
>> > being calculated in the above situations? (in loose mathamatical
terms if
>> > possible, as you can probably tell from my question).
>>
>> In loose terms the correlation measures the variability between
blocks
>> relative to the variation
>> within blocks. Over-simplifying the design matrix will increase
the
>> between-blocks variation,
>> because it will now reflect differences between your treatments as
well as
>> differences between
>> biological replicates. Hence the estimated correlation increases.
>>
>
> Okay. Now I believe I understand how it is calculated. When you use
a design
> matrix here you create blocks, then the blocking argument creates
blocks within
> blocks. (Correct me if this is wrong).
>
> Best Regards, Carolyn